Will Apple's G5 come from IBM? - Page 2

162 instructions in AltiVec? I thought it was 160...

<a href="http://developer.apple.com/hardware/ve/detail.html" target="_blank">http://developer.apple.com/hardware/ve/detail.html</a>

EDIT: added URL

[ 08-08-2002: Message edited by: Eugene ]</p>

[quote]Originally posted by sc_markt:
<strong>

Easy for you to say, you probably have a G4. I've had this 8600 for 5 years. I can't relax. And, I can't run OS X.</strong><hr></blockquote>

Yeah, you can....well kinda slowly, but I'm running it on my 9600/350, and it's actually quicker than it was on my Wallstreet.

I don't think IBM will supply their Power4 chips for the Mac. The key issue here is the cost. Power requirements and heat dissipation will also lead to design problems. Even if they license Altivec from Motorola, I don't think Apple can afford to use it. The best bet is still on Motorola. It seems to me clockspeed are the not major factor of concern but rather, that turtle system bus draws the most concern. No matter what Altivec can do, if the bus is slow, the processor will just have to sit there for those endless cycle before it can do anything. Process speed to bus ratio has lengthed to 7.5 to 1 for Apollo. So lets say we continue to get speed bump, the processor will have to wait longer to get the data in. What a waste? Those L2 and L3 are great but they are data that are frequently used and they still have to be fed by the system bus at the beginning. Now graphics hardware demands more, and HD and other connecting buses such as Firwire and USB are getting speed boost, if system bus doesn't get a boost, the performance boost to the overall system will be severely limited to the system bus itself. If thats the case, forget AGP 8X and Gigawire and USB2 because that turtle bus can't keep up, so even if Moto continues to keep the same pipeline depth, and rev it up to 2 GHZ, your Mac will still be a turtle. What the promac needs is a faster bus, and slight increase in pipeline depth. I think Motorola can do that, they already rolled out G5 embedded solution. G5 main architecture change according to the roadmap is a die shrink and new bus, couple with new pipelines. I believe that is what Steve is going to roll out in a week time.

[quote]Originally posted by cowerd:
<strong>
My guess is that they will still be low-power, or relatively low-power. Entry-level servers == blade servers, where heat dissapation and power consumption are a concern. Not as much of a concern as embedded, but you won't be cooking eggs on the thing.</strong><hr></blockquote>

Well, if they can make an 8 way processor that "breaks through compute intensive barriers" low-power as well, more power to them, I say. But I don't think so.

The first thing IBM, or any other embedded processor company really, does to make low-power chips is to cut down the number of execution units. I've yet to hear of an embedded processor with two FPUs. It looks like they're going all-out for performance on this one. 'Bout time!!

[quote]I don't think IBM will supply their Power4 chips for the Mac. The key issue here is the cost. Power requirements and heat dissipation will also lead to design problems. Even if they license Altivec from Motorola, I don't think Apple can afford to use it. <hr></blockquote>

Uhh... The processor being discussed here is not the Power4, it's a PPC based on the Power4, geared towards "desktops and entry level servers". Certainly sounds like Apple might be interested in that, don't you think?

[quote]Originally posted by sizzle chest:
<strong>

Actually, there have been several machines between the 8600 and these next-next-next generation ones, that will run OSX. You could have purchased one of those. And if those aren't good enough, buy one of the new ones coming out in the next month or so.

There will ALWAYS be bleeding-edge, early design stage stuff we'll hear about, that's faster than anything we can actually buy. It doesn't mean the stuff we can buy isn't worth buying, just because the bleeding-edge stuff is coming eventually.</strong><hr></blockquote>

I'm in a better position than sc, since in addition to my 8600, I have an iBook 466. However, my 8600 is still my main machine, mainly due to screen size, so I feel for sc. My 8600 is going to be five years old, and I would love to have OS X on the desktop. And while there are many capable machines out there one could purchase, why not look instead at the logic of the situation?

As another poster opined, anyone who bought a G4 early on or at the half way point in its progression, they made a good purchase. The reason is the G4 didn't progress all that far - an increase of 500MHz with a loss of IPC efficiency. Now when I bought my 8600 to replace my Quadra 650, my 8600 had a more efficient processor with nearly ten times the clock speed. If Apple had made the same strides with the G4 that it made between my 650 and my 8600, we would have multi-GHz G4s right now.

Yet, as the MHz gap turned into a GHz gap, Mac users woke to the unfortunate truth that the G4 is woefully inadequate. Now IBM has just announced a chip that promises to blow the G4 away, a leap that should even dwarf the comparative difference between my 650 and 8600. Realize we're not simply talking about moving from a G4 @1000MHz to a G4 @1200MHz (which is probably all we'll get in the short term). We're talking instead about a huge leap in technology. The G4 has been holding Apple back; it will be dwarfed by this modern IBM chip. With this in mind, who could contemplate buying one of Apple's current desktop offerings? If Apple is going to use this new IBM line (and that's probably the only plausible inference to draw), then I'm waiting for the new POWER Macs, even if they're another year off or beyond.

[quote]Originally posted by tiramisubomb:
<strong>I don't think IBM will supply their Power4 chips for the Mac. The key issue here is the cost. Power requirements and heat dissipation will also lead to design problems.

[SNIP]</strong><hr></blockquote>

I think you may have overlooked the basis of this particular thread. From what we know now, I believe the IBM G5 is practically guaranteed.

[ 08-08-2002: Message edited by: Big Mac ]</p>

Well, if G5 is a version based on Power 4/5, it will be very interesting, but I still think G5 is probably going to come from Motorola. I believe both IBM and Motorola can offer a build to order option for Apple so that SJ can choose what features to add on.

[quote]Originally posted by tiramisubomb:
<strong>Well, if G5 is a version based on Power 4/5, it will be very interesting, but I still think G5 is probably going to come from Motorola. I believe both IBM and Motorola can offer a build to order option for Apple so that SJ can choose what features to add on.</strong><hr></blockquote>

We just had a breakthrough announcement about new 64 bit IBM PPCs aimed at the desktop and entry server markets, based on the Power4 with a vector processor that sounds remarkably like Altivec. On the other hand, the only thing Motorola has announced concerning the G5 is its embedded line. There's nothing currently announced from Motorola even hinting at a desktop G5. To the contrary, we've heard rumblings that Apple axed any G5 plans it had with Motorola. I'm certainly sympathetic to the skeptic view, but I think it's time to go with the preponderance of the evidence.

[ 08-08-2002: Message edited by: Big Mac ]</p>

"Speaking of which, Buh Bye Moto-scro-la. Rat bastards."

Why...Moogs...you sound...almost...bitter?

So it seems IBM will give us the true 'G5' next generation PPC. That much looks obvious now. The question is 'when'. It looks like it could be this time next year, folks.

I think we'll just get a '1.8 G4' on Rio in the meantime...for Macworld San Francisco.

This chip will be 7470 stretched even further or the 7500? Probably the latter my guess.

Which won't be a bad performer...it will just have the crap beaten out of it by IBM's 'G5'.

Now? 7470 1.4 G4s on DDR/BUS/Mobo...

Moto's G4s will filter into the consumer line over the next year or so.

Not a bad future then. Good enough.

Lemon Bon Bon

Let me be the first to congratulate Andrew Welch (moki) for being totally right.

[quote]Originally posted by sizzle chest:
<strong>

Where does it say "velocity engine?" It does say "vector unit," and Altivec/Velocity Engine IS a vector unit. Altivec/Velocity Engine is a vector unit, but not all vector units are Altivec/Velocity Engine.</strong><hr></blockquote>

Take a wild guess as to what this means. It's as simple as adding 1+1.

BTW, "Velocity Engine" is Apple's term for the instruction set. Altivec is Motorola's term for their implementation of the instruction set. When IBM implements the same thing, Apple can still use Velocity Engine.

[quote]Originally posted by tiramisubomb:
<strong>I don't think IBM will supply their Power4 chips for the Mac. The key issue here is the cost. Power requirements and heat dissipation will also lead to design problems. Even if they license Altivec from Motorola, I don't think Apple can afford to use it. &lt;snip&gt;.</strong><hr></blockquote>

Others have pointed out: this is not a power4, but a powerpc based on a power4.

Also, I believe Apple has full (free) rights to the velocity engine instuction set. They would be stupid not to. Motorola probably has rights to their implementation of that instruction set (Altivec), but there's nothing preventing Apple from having IBM implement the same thing.

[quote]Originally posted by Aris:
<strong>"question about the nforce2 chipset.
CAN it even be used on a NON x86 architecture?</strong><hr></blockquote>

No. nVidia would have to do a special version for Apple. But $$$ is a big motivator.

However, if you take the Macintosh reference from the nforce2 description and combine it with the statement from nVidia's CEO about the nVidia graphics on Macs, it adds up to a good probability that there's a partnership here. Using the nforce2 (one implementation anyway) would put a gforce4mx on every mac. I know most others discard this nforce rumor stuff, but I'll stake my reputation as a newbie to this forum that it's accurate.

[quote]Originally posted by tiramisubomb:
<strong>I believe both IBM and Motorola can offer a build to order option for Apple so that SJ can choose what features to add on.</strong><hr></blockquote>

Never happen. Neither IBM nor Motorola would ever develop a new processor in the hopes that Apple would use it. Speed bumps, yes. A new processor which would mean a significant amount of reworking at Apple? No. This isn't like the peecee market where Intel and AMD can come out with a new processor and be guaranteed that enough companies will buy it to cover the development cost.

For something like this, Apple is probably giving them a guarantee of purchasing a certain amount over a certain time. Either that or theyre picking up the tab for a big chunk of the development costs.

I don't understand why there should be a difference between IBM's chip being "a Power4", and it being "a PPC based on Power4" - after all, Power4's two kernels ARE PPC kernels, right?

What are the technical differences?

engpjp

[ 08-08-2002: Message edited by: engpjp ]</p>

[quote]Originally posted by engpjp:
<strong>I don't understand why there should be a difference between IBM's chip being "a Power4", and it being "a PPC based on Power4" - after all, Power4's two kernels ARE PPC kernels, right?

What are the technical differences?

engpjp

[ 08-08-2002: Message edited by: engpjp ]</strong><hr></blockquote>


PowerPC is a subfamily of the POWER CPU and Current PPC G3-750FX and G4 are based on POWER 3, if I remember correctly.

Aw

It's all a little too nicely wraped up, and pretty for me. I'm not going to bite.
I've seen Mac fanatics go ape, and even become hostile before when their predicted imaginary products don't show up. Then they have the balls to blame everybody accept themselvs. And they say stuff like "Where's my Power 4 derived PPC G5?"

As if?

[quote]Originally posted by engpjp:
<strong>I don't understand why there should be a difference between IBM's chip being "a Power4", and it being "a PPC based on Power4" - after all, Power4's two kernels ARE PPC kernels, right?

What are the technical differences?

engpjp

[ 08-08-2002: Message edited by: engpjp ]</strong><hr></blockquote>

As IBM hasn't released the technical details of this new chip yet, one can only speculate about the specifics.

However, others have correctly pointed out that Apple is unlikely to use the Power4 because of its cost and heat. It has HUGE amounts of cache that would be nice, but the average consumer wouldn't be willing to pay for.

Another way to look at it might be Power4 lite + velocity engine.

[quote]Originally posted by Cat:
<strong>It's all a little too nicely wraped up, and pretty for me. I'm not going to bite.
I've seen Mac fanatics go ape, and even become hostile before when their predicted imaginary products don't show up. Then they have the balls to blame everybody accept themselvs. And they say stuff like "Where's my Power 4 derived PPC G5?"
</strong><hr></blockquote>

You know, I don't think Mac fanatics care one way or another whether it has a Power4 or any other specific piece of hardware. What they're looking for is SPEED. Lots and lots and lots of SPEED. They don't care if that speed comes from a Power4, a G4+++, a G3+++++++, nVidia, AMD, or whatever. (With the possible exception of anything Intel.) Give them SPEED and they'll be happy.

Seeing these little trickles of technical info simply give them hope that SPEED is just around the corner.

Well, IBM has NOT been any great contributor to the PPC. They ***NEVER*** had any G3 running faster than equivalent G4's over the past 2 and a half years.

They talked up a nice game, but they didn't produce anything. If you go back they originally talked about Sahara **with a VPU** starting at 1Ghz and it was to already have reached 2Ghz. Then the actual chip arrived and it had no VPU and it started at 600Mhz and only reaches 800Mhz, with 1Ghz coming? VPU and 2GHz??? The story/projections then changed to say we can expect those things in LATE 2003!!!

SO, given IBM's track record, even if they start talking about a 64bit VPU equipped PPC, it might be as much as 2 years before you see it.

Don't get too excited too fast. As far as PPC goes IBM has sucked siphilitic monkey balls even harder than Moto has.

[ 08-08-2002: Message edited by: Matsu ]</p>

[quote]Originally posted by Appleworm:
<strong>


PowerPC is a subfamily of the POWER CPU and Current PPC G3-750FX and G4 are based on POWER 3, if I remember correctly.

Aw </strong><hr></blockquote>

The differences between the earlier POWER chips and the PowerPC are all gone. POWER 4 is a 64 bit PPC implementation. PPC is a 64 bit instruction set with a 32 bit subset which the 32 bit chips use. POWER 4 is %100 compatible with 32 bit PPC.

check out IBMs roadmap.

<a href="http://www-3.ibm.com/chips/products/powerpc/rdmap/roadmap_small.jpg" target="_blank">http://www-3.ibm.com/chips/products/powerpc/rdmap/roadmap_small.jpg</a>

and check out an IBM article on Power and PowerPC to see that PowerPC is a derivative of Power and yet very similar.

<a href="http://www-1.ibm.com/servers/eserver/pseries/hardware/whitepapers/power/ppc_arch.html" target="_blank">http://www-1.ibm.com/servers/eserver/pseries/hardware/whitepapers/power/ppc_arch.html</a>

[ 08-08-2002: Message edited by: Slacker ]</p>

[quote]Originally posted by Slacker:
<strong>check out IBMs roadmap.

<a href="http://www-3.ibm.com/chips/products/powerpc/rdmap/roadmap_small.jpg" target="_blank">http://www-3.ibm.com/chips/products/powerpc/rdmap/roadmap_small.jpg</a></strong><hr></blockquote>

The thing I hate about these roadmaps is the timeline. 1999-200X might mean 2003, but could mean 2009. So we have no real way of knowing when this 2GHz beast will show up, except for the fact that IBM has it's .13 process going and a even smaller one in development.

It's all too vague.

[quote] Originally posted by Matsu:

They ***NEVER*** had any G3 running faster than equivalent G4's over the past 2 and a half years. <hr></blockquote>

I had assumed that the reason for this was due to the Apple marketing perspective (i.e. there is no way that they could have a faster clockspeed G3 than their flagship G4 chip in any of their 'consumer' computers)?

[Edit; Maybe I misunderstand what you mean by 'equivalent']

[ 08-08-2002: Message edited by: DaveLee ]</p>

[quote]Originally posted by Blackcat:
<strong>

The thing I hate about these roadmaps is the timeline. 1999-200X might mean 2003, but could mean 2009. So we have no real way of knowing when this 2GHz beast will show up, except for the fact that IBM has it's .13 process going and a even smaller one in development.

It's all too vague.</strong><hr></blockquote>

100 % Agree
I think the 2 GHz+ will be shipping 2006-2007...


Aw

[quote]Originally posted by Slacker:
<strong>check out IBMs roadmap.</strong><hr></blockquote>
Pay particular attention to the fine print in the bottom right corner.

*subject to change without notice

Never happen. Neither IBM nor Motorola would ever develop a new processor in the hopes that Apple would use it. Speed bumps, yes. A new processor which would mean a significant amount of reworking at Apple? No. This isn't like the peecee market where Intel and AMD can come out with a new processor and be guaranteed that enough companies will buy it to cover the development cost.

For something like this, Apple is probably giving them a guarantee of purchasing a certain amount over a certain time. Either that or theyre picking up the tab for a big chunk of the development costs.
------------------------------------------------------------------------

Well, what I really mean is that Motorola and IBM can both offer their unique features that will benefit Apple. Let me put it in example, the current Apollo has 1 GHZ clockspeed, but Motorola did not do much for the system bus or they simply refuse to enhance it just for Apple. However, the PowerPC Book E architecture allows IBM and Motorola chips to share common architecture. And in this case, if IBM has a better bus and other elements that can benefit Apple, Apple could just request so that both Motorola and IBM elements of the chips can be put together for Apple. Think of it as Built to Order concept. The architecture of the chip was well layout, now the customers can choose what to add.

[quote]Originally posted by DaveLee:
<strong>

I had assumed that the reason for this was due to the Apple marketing perspective (i.e. there is no way that they could have a faster clockspeed G3 than their flagship G4 chip in any of their 'consumer' computers)?

[Edit; Maybe I misunderstand what you mean by 'equivalent']

[ 08-08-2002: Message edited by: DaveLee ]</strong><hr></blockquote>


And I also think that IBM didn't want to spend too many time to work with Apple because they choose moto... but moto being a dead-end, IBM-Apple processors will have a nice roadmap...

Aw

[ 08-08-2002: Message edited by: Appleworm ]</p>

Wow, I can't believe all the crap being spewed in here! C'mon guys you're generating crazy amounts of FUD for no good reason.

- IBM has been a huge contributor to the PowerPC movement. They invented the POWER architecture on which it is very closely based, after all.
- The POWER architecture internal design diverged from the PowerPC starting with the POWER2 and they have been unrelated (internally) since then. They have both been very close to compatible, however, and with the POWER4 they are fully compatible at the user ISA level.
- The G3/G4 are not in any way connected to the design of the POWER3.
- IBM decided not to include VMX in their chips because they didn't believe it was the correct way to build high performance processors. Plausible argument at the time, it has since been proven wrong and they acknowledged that a couple of years ago.
- IBM took part in the VMX design process, and are co-listed on the couple of patents I've seen. My guess is that they can't use Moto's implementation, but are free to build their own compatible implementation.
- The POWER4 chip is ~170 million transistors, however that includes two cores and a huge pile of cache. Each core is purportedly about 30 million transistors, which is of the same complexity level as the Pentium4 and Athlon. It works on the PC, it'll work on the PowerPC.
- Architecturally it is very different internally from the G1/G2/G3/G4 in that it "cracks" instructions into internal operations which can then be dealt with by a very highly pipelined and superscalar machine. This ought to allow for very high clock rates (i.e. 2GHz+) with many execution units. Not much point in having an 8-way superscalar machine if you don't have the execution units to feed.
- Most of the believeable sources point to a mid-to-late 2003 introduction for this processor, and this fits with moving to a 0.1 micro (or smaller) process. Apple isn't going to curl up and die in the intervening 12-16 months, especially if the 1.4 - 1.8 GHz G4s come from Motorola and there are solutions to the bandwidth problem.
- The new processor will certainly be 64-bit so Apple needs to get a 64-bit version of OSX up and running by then. Hmmm... that would mean they could build a machine based on the Hammer too (HyperTransport interface and all).
- The 7xx series won't necessarily stop where it is. That line has a different purpose -- low-power, low-cost. I could easily seem them adding VMX and continuing to target the iBook.

Guys be positive. Geez the signs are hitting you in the face .

IBM announces new Chip Foundry

Oh suprise they also announce a new Power4 derivative for Desktops with a Vector Unit no less.

IBM is not Motorola...don't get them confused.

You know more today about the potential for staying PPC than you did just two days ago.

I'm happy. I'm not interested in X86 Macs. However I do believe the current G4 may have alot of life yet. I doubt you see a Power4 Derivative in a notebook anytime soon.

Peter Sandon from IBM who will be giving the presentation on the new chip is the same guy who authored the papers / releases on the 750CX/FX which in my opinion makes this highly credible as a replacement to the G4 on high end machines. Peter is the Senior PowerPC Processor Architect.

I would guess it would have to be an early 2003 release if it really is a replacement for the G4. Once this chip is confirmed as a part apple could use, Apple's high end sales will go stagnant until it's release. (Not that they've been great in recent months anyways).

I think we have the real deal here.

[quote] I'm happy. <hr></blockquote>

Me too

[quote] Once this chip is confirmed as a part apple could use, Apple's high end sales will go stagnant until it's release. <hr></blockquote>

I think Apple will have to realise soon that they HAVE to make an announcement of some form or other with regard to where their architecture is going. There has been too much rampant speculation about what Apple must/will do, among others this includes much of the 'legitimate' press (i.e. Macworld magazine) and these 'analysts' that are beginning to make waves. Apple really should clarify their commitment to PowerPC and yield some information about what we can expect from them in the future. Jeez, they're not dealing with national security here.

[quote]Originally posted by Programmer:
[QB
- The POWER4 chip is ~170 million transistors, however that includes two cores and a huge pile of cache. Each core is purportedly about 30 million transistors, which is of the same complexity level as the Pentium4 and Athlon. It works on the PC, it'll work on the PowerPC.
.[/QB]<hr></blockquote>

Considering that the new chip will have only one core, and less cache (i consider that they will remove the L3 cache and limit the L2 cache between 512 k and 1 MB) : i say 30 MB for the core, add 10 M transistor for the altivec, and 10 millions more for the cache : it make 50 millions of transistors, like the Programmer said the same amount of transistors that the last release of pentium 4. Considering that the derived chip, let's call him the power VMX is built on SOI 0,13 , it will make less heat thant the pentium 4 on 0,13 (but without SOI).
There is no reason why such a chip can't work in a Apple. Considering that the latter version of the tower have a more big (efficient ventilation) the power VMX will be perfect in it.

Here's my bet we will see a power VMX next year : if i win i will buy one

[quote]Originally posted by Programmer:
<strong>IBM took part in the VMX design process, and are co-listed on the couple of patents I've seen. My guess is that they can't use Moto's implementation, but are free to build their own compatible implementation.</strong><hr></blockquote>

As a point of interest, after I saw those 2 vector patents that named IBM, Motorola, and Apple jointly as assignees (filed in '98), I did a search on the named inventors (which I would assume were employees from all 3 companies).
I came up with about 30 more patents (many filed in 2000),mostly vector related, all SIMD related, all with the exact same inventor names, and all filed in IBM's name alone.
Kinda made me wonder if those employees have all moved to IBM, or if IBM is filing (under legal contract showing joint ownership) for all to keep the competition from knowing what's up.

[quote]Originally posted by hmurchison:
<strong>I'm happy. I'm not interested in X86 Macs. However I do believe the current G4 may have alot of life yet. I doubt you see a Power4 Derivative in a notebook anytime soon.</strong><hr></blockquote>

Hm, that triggered a thought. Will G3s be left in the iBook? Or will there be a clean split:
G4s for portables (and the iMac)
Power4 derivatives for Power Macs

Maybe that's how Moto stays in the game -- with a demotion.

Screed

[ 08-08-2002: Message edited by: sCreeD ]</p>

[quote]Originally posted by Powerdoc:
<strong>

Considering that the new chip will have only one core, and less cache (i consider that they will remove the L3 cache and limit the L2 cache between 512 k and 1 MB) : i say 30 MB for the core, add 10 M transistor for the altivec, and 10 millions more for the cache : it make 50 millions of transistors, like the Programmer said the same amount of transistors that the last release of pentium 4. Considering that the derived chip, let's call him the power VMX is built on SOI 0,13 , it will make less heat thant the pentium 4 on 0,13 (but without SOI).
There is no reason why such a chip can't work in a Apple. Considering that the latter version of the tower have a more big (efficient ventilation) the power VMX will be perfect in it.

Here's my bet we will see a power VMX next year : if i win i will buy one </strong><hr></blockquote>

Just for speculation, how about 80 million transistors - dual core with AltiVec
Maybe on a smaller proccess?

What is the current transistor count in the G4?
It's sure a lot smaller (in physical size) than a P4. So, is this a transistor count difference, or a something else difference?

[ 08-08-2002: Message edited by: taboo ]</p>

[quote]Originally posted by Matsu:
<strong>Well, IBM has NOT been any great contributor to the PPC.</strong><hr></blockquote>

Well so far as the Apple crowd is concerned. They have done massive amounts for their own desktops/workstations/servers.

I think I am going to make a big spending next year